IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO.

2, APRIL 2005 363

Complex State-Space Modeling and Nonlinear

Control of Active Front-End Converters
Rolando P. Burgos, Member, IEEE, Eduardo P. Wiechmann, Senior Member, IEEE, and Joachim Holtz, Fellow, IEEE

Abstract—This paper presents the modeling and control of in these applications, where these rectifiers have become known
active front-end (AFE) converters using complex state-space as active front-end (AFE) converters.
representation, a technique developed and thus far mostly em- All these appealing characteristics have been achieved thanks
ployed for the analysis of ac machines. Particularly, three-phase
PWM voltage-source and current-source rectifiers are thor- to the development of high-power semiconductor devices and
oughly studied using the graphical capabilities of this approach, advanced control strategies [2], [7]–[9], upon which state-feed-
namely, complex signal flow graphs. These are used to directly back input–output linearization (IOL) has been one of the
and intuitively derive high-performance nonlinear control laws most promising [10]. The main achievement of this nonlinear
based on input–output feedback linearization. Specifically, a
cascaded and a paralleled control scheme are investigated for the
control technique has been the decoupling and linearization
voltage-source rectifier, whereas a cascaded scheme is considered of the synchronous - axes dynamics in AFE converters [5],
for the current-source rectifier. Under these strategies both con- [11]–[13], which furthermore ensures the stability of high-per-
verters exhibit linear and decoupled – axes dynamics, while formance regenerative drives [14]. The attractiveness of IOL
also attaining a reactive power compensation capacity. Moreover, has not passed unattended. On the contrary, it has been applied
linearization of their respective dc-link voltage and current loops
utterly enforces and ensures their operating stability. All this with excellent results to motor drives [15]–[17], reactive power
is achieved without the elaborate mathematical complexity of compensators [18], and power-factor correction circuits among
input–output linearization, effectively shunned out by the pro- others [19]. However, its involved mathematical intricateness
posed complex state-space approach. Finally, experimental results has somehow constrained the widespread usage of this tech-
from 5-kVA digital-signal-processor-based laboratory prototypes nique, so that simple linear approaches —yet having nonlinear
verify the analysis and downright performance evinced by these
AFE converters. dynamics—are still preferred [3], [7]–[9].
On the load side of these AFE converters there is also a signif-
Index Terms—Active front-end (AFE) converter, complex
state-space, input–output linearization, pulsewidth-modulated icantly elaborate system, the electrical machine, whose differen-
(PWM) current-source rectifier, pulsewidth-modulated (PWM) tial equations represent a highly coupled nonlinear system [20].
voltage-source rectifier. Notwithstanding, a simple yet effective way of modeling their
dynamics has enlightened their operation by employing com-
plex state variables and space vector theory [21]. The unique-
I. INTRODUCTION
ness of this approach is that it enables the modeling, analysis,

T HE decrease of generation-consumption capacity margins

in electrical power systems around the world has impelled
the development of ever stringent, harsher standards and regula-
and control of ac machines with a single tool. In effect, by using
complex signal flow graphs it renders systems of complex state
variable equations intelligible by visual inspection, providing
tions [1], [2]. These have furnished the development and usage meaningful information both for analysis and control purposes.
of higher efficiency static power converters, and especially This method has been successfully employed for studying ma-
so of those which directly interact with the power system. chine electromagnetic phenomena [22], [23], and for developing
Three-phase pulsewidth-modulated (PWM) voltage- and cur- advanced control strategies for them [24], [25]. However, in
rent-source rectifiers (VSRs and CSRs) have consequently spite of its numerous advantages this approach remains virtu-
emerged to meet these taut requirements for motor-drive ap- ally unexploited for other applications.
plications [3]–[6]. Characteristics such as sinusoidal input This paper presents thence the modeling and control of
currents, controllable input power factor, high-quality dc-link AFE converters using complex state-space representation, thus
variables, and regenerative capability are now common features expanding the usage of this particular technique. Specifically,
three-phase PWM-VSRs and CSRs are fully analyzed using
Manuscript received December 30, 2003; revised February 9, 2004. Abstract complex signal flow graphs, from where nonlinear control
published on the Internet January 13, 2005. This work was supported by laws are directly obtained based on IOL principles [10]. Both
FONDECYT (Chilean Fund for Science and Technology Development) under
Project 104-0473. This work also made use of Engineering Research Centers
cascaded and paralleled current/voltage control structures are
Shared Facilities supported by the National Science Foundation under Award assessed for the PWM-VSR [11], [12], [26], [27], whereas a
EEC-9731677. cascaded input-current/dc-link current scheme is presented for
R. P. Burgos is with the Center for Power Electronics Systems, Virginia
Polytechnic Institute and State University, Blacksburg, VA 24060 USA (e-mail: the PWM-CSR [5], [14]. The intuitiveness of this approach
rp.burgos@ieee.org). becomes apparent from the analyzes performed. As a result, the
E. P. Wiechmann is with the Electrical Engineering Department, University AFE converters under study attained decoupled and linear -
of Concepción, Concepción, Chile.
J. Holtz is with Wuppertal University, 42097 Wuppertal, Germany. axes dynamics, featuring an unaltered dynamic response with
Digital Object Identifier 10.1109/TIE.2005.843919 enforced stability regardless of the converter operating point.
0278-0046/$20.00 © 2005 IEEE
364 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

The usefulness of this method has already been established by

aiding in the design of more complex systems, such as PWM ..
.
rectifiers for medium voltage current-source drives [28].
This work includes a brief theoretical description of IOL,
the complete modeling of the AFE converters using complex
.. .. ..
state-space notation, together with an in-depth description of . . .
the nonlinear control laws derived. Experimental results from
5-kVA TMS320C32 digital-signal-processor (DSP)-based lab-
oratory prototypes are presented and used for validation of the ..
. (5)
proposed approach and strategies.

II. IOL or equivalently

A brief theoretical background of IOL is herein presented for
the sake of completeness of this work; a detailed description of
it might be found in [10]. Consider the following multiple-input .. (6)
.
multiple-output (MIMO) nonlinear system,

Then, if is invertible, the input transformation

(1)
(7)
where is an state vector, is an output vector,
yields equations of the form
and are th-order smooth vector fields, is
an matrix of smooth vector field columns , and is
(8)
the control input vector. The essence of IOL is to generate
a linear differential relation between the desired system outputs which is a linear differential relation between each input–output
and a set of new control inputs , which correspondingly pair and . Moreover, since each input only affects its
supersede the original system inputs . correspondent output , (7) is also termed a decoupling control
IOL is obtained as follows. Firstly, outputs are differenti- law.
ated until at least one of the inputs appears. The smallest number
of derivatives required for that, , is called the relative degree
III. PWM-VSR TOPOLOGY AND MODELING
of that particular output. These derivatives may be expressed as
follows: A. Stationary Frame State-Space Model
Fig. 1 shows the schematic of a PWM-VSR employed for in-
(2) duction motor drives, where is the input current to another
voltage-source inverter that feeds a motor load. The average
converter input–output electric relationships may be written per
the transfer function concept as follows [29]:
where for at least one in (2) if the previously
mentioned condition is complied. Equation (2) is built using Lie
(9)
derivatives, scalar functions of the following form:
(10)
(3)
where is the converter ac phase-voltage vector
is the modulating or control input vector
where is a smooth vector field on is a smooth scalar and are the dc-link voltage and current,
function, and is the gradient operator. As seen, these are di- respectively, and is the input line current vector .
rectional derivatives of along vector . They may also be de- The state-space model of the converter is then given by
fined recursively as shown below

(11)
(4)
(12)
Continuing, if (2) is complied with for every output, the fol-
lowing system of differential equations is obtained:
where is the power supply phase-voltage vector
is the input filter inductance, is the dc-link
capacitance, and and are parasitic resistances of the
..
. inductors and capacitors not shown in Fig. 1 for the sake of
clarity. The dc-link voltage dynamics may be alternatively
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 365

Fig. 1. PWM-VSR AFE converter schematic.

expressed in terms of the converter input–output power balance Then, by applying (15) and (17) to (11) and (12), the fol-
in accordence to the following relationship between its ac and lowing complex state-space system is obtained:
dc sides:
(18)
(13)
(19)
This expression assumes that the losses on both ac inductors and
power converter are negligible. Finally, rearranging (13) yields
where time constants and are, respectively, defined as
the dynamics of the dc-link voltage in its state-space form as
and , and the complex-variable inner product used
shown below
in (19) by
(14)
(20)

B. Synchronous – Axes Complex State-Space Model Alternatively, the dynamics of may be expressed in terms of
The following transformation is used in this paper to trans- the converter power balance as follows:
form abc-frame variables into space vectors:
(21)
(15)

where , and are three-phase variables, and is IV. PWM-CSR TOPOLOGY AND MODELING
the resultant space vector expressed as follows in the complex
plane: A. Stationary Frame State-Space Model
Fig. 2 shows the schematic of a PWM-CSR, where rep-
(16) resents the current-source inverter and motor load. Accordingly,
the average input–output converter relationships are defined by
Hereinafter, complex variables will be written in bold fonts.
Equation (15) represents a transformation into the stationary (22)
– frame. A rotating transformation into the synchronous –
axes frame might then be defined as follows: (23)

(17) where is the converter ac phase-voltage vector

is the modulating or control input vector
where is a unity space vector rotating clockwise at syn- and are the dc-link voltage and cur-
chronous speed . rent, respectively, and is the input line current vector
366 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

Fig. 2. PWM-CSR AFE converter schematic.

. The state-space model of the converter is then Its power balance relation may also be converted into the com-
given by plex state-space as shown below

(24) (32)

(25) In (32) time constants , , and are, respectively, defined

as
(26)
(33)
where is the line current vector is the
power supply phase-voltage vector is the input
filter inductance, is the dc-link filter inductance, is the V. CASCADED REPRESENTATION AND CONTROL
input filter capacitance, and , and are parasitic resis- OF THE PWM-VSR
tances of the inductors and capacitors not shown in Fig. 2 for A. Complex Signal Flow Graph
the sake of clarity. The dc-link current dynamics may also be
expressed in terms of the converter input–output power balance The PWM-VSR may be represented graphically by means of a
given below complex signal flow graph, which is an equivalent representation
of (18) and (19). Fig. 3(a) shows the resulting diagram comprised
(27) of three blocks interconnected by signal flow lines, each of
these modeling the ac inductors, dc-link capacitor, and the
The assumptions in this case are negligible losses in the input converter itself. The inputs to this system are the power supply
– filter and the converter itself. Rearranging (27) yields its voltage , the inverter load current , and the control
state-space form as shown below input , which, respectively, excite the inductors, capacitor,
and converter blocks. The dynamics of the reactive elements
(28) are represented by first-order complex time delays of time
constants and . The inductors blocks are also excited by two
B. Synchronous – Axes Frame Complex State-Space Model additional negative feedbacks, the cross-coupled inductors back
electromotive force (EMF) (only observed from a nonstationary
A complex state-space model of the PWM-CSR may be ob- frame, ), and by the converter ac voltage , which
tained by applying (15) and (17) to (24)–(26), which yields the evinces its nonlinear nature. The output of the inductors block
following system of complex differential equations: is the ac current , which in turn excites the capacitor block
through its nonlinear interaction with the converter that produces
(29) the dc-link current . The output of the capacitor block, ,
flows back then through the converter to close the dynamic
(30) interaction of the PWM-VSR state variables. The scalar nature
of the capacitor block is illustrated by the use of single-line
(31) flow arrows.
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 367

Fig. 3. Complex signal flow graph of the cascaded PWM-VSR representation. (a) Input current i and dc-link voltage v dynamics. (b) Converter input–output
power balance-based dc-link voltage dynamics.

The input–output power balance of the PWM-VSR cannot be control inputs that are bound to replace . Such a relationship,
directly analyzed from Fig. 3(a), as this signal flow graph solely for instance, is
represents the current and voltage dynamics. For that purpose,
a separate signal flow diagram equivalent to (21) is developed (34)
and shown in Fig. 3(b). This is an alternative representation to
study the dc-link voltage dynamics, which nonetheless are also The advantage of the proposed approach is that the current
captured by means of a first-order complex time delay of con- control law may be simply and directly obtained by inspection
stant . Its inputs in this case however are the power supply of the complex signal flow graph, as opposed to following the
voltage and ac current , whose product represents the con- involved mathematical procedure demanded by IOL. In fact, by
verter input power, and the load current representing the observing in Fig. 3(a) the signals and terms that need to be can-
output power of the converter. As expected, the instantaneous celled out to comply with (34), and by equating this same equa-
difference between the input and output power charges or dis- tion with (18), the input transformation may, thus, be obtained.
charges the dc-link capacitor. Notice that in this case current This current control law is shown below
is replaced by its reference , which corresponds to the input
control signal of this power system representation. (35)

The structure of this input transformation clearly resembles that

B. Inner Input-Current Loop of (7), having both a linearizing term and a decou-
pling term . In this way, IOL of the
According to IOL, a control law or transformation of the con- PWM-VSR current has been attained by simple feedback lin-
trol input must be defined in order to establish a linear differ- earization thanks to the usage of complex signal flow graphs.
ential relationship between the system outputs and the new sets Finally, in order to tune the dynamic response of this loop,
of control inputs , as shown in (7). This can be done following the new control input is defined in terms of the current error
the IOL aforementioned procedure, or intuitively by using com- as follows:
plex signal flow graphs as will be herein presented.
By examining Fig. 3(a), a first-order dependence of current (36)
is observed from the control input , as product is one
the three signals directly exciting the inductors first-order time where in (36) , and constants
delay block. This suggests that a first-order linear differential and are used for pole placement of the desired transfer
relationship may be constructed between and the new set of function between the Laplace-transformed variables and
368 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

Fig. 4. Block diagram of the cascaded control scheme for the PWM-VSR. (a) Input current i loop. (b) Outer dc-link voltage v and reactive power q loops.

according to settling time . Their specific values are actual control signal of the outer loop. This again implies that a
determined per the ITAE criterion (integral of time multiplied first-order dependence exists between and its control signal,
by the absolute value of error) optimized for a ramp input as also dictating the type of linear differential relationship that may
follows [30]: be constructed between and its new control signal , which
is explicitly shown below
(37)
(38)
This linear controller features exponentially convergent
tracking dynamics—which greatly enhances the performance Then, by equating (21) and (38) and by observing the signals
of this inner loop—attained by adding the time derivative of and terms that need to be cancelled out in Fig. 3(b), the dc-link
. This does not increase the noise sensitivity of the control voltage input transformation may be written as follows:
law as is defined and bounded by the outer loops. The
controller also presents parameter mismatch compensation (39)
through its integrator term. This ensures zero steady-state error
even when subject to temperature or aging parameter variations In this control law, has been nulled through proper synchro-
given the nature of the nonlinear transformation, which depends nization of the space-vector transformation, hence, may be
on system and converter parameters as shown in (35). Fig. 4(a) solely controlled by the -axis current. The new control signal
shows the resulting block diagram for this current control loop. is then defined in terms of the dc-link voltage error , and
used to tune the dynamic response of this loop as shown below
C. Outer DC-Link Voltage Loop
The complex signal flow graph depicted in Fig. 3(a) gives (40)
the false impression that a first-order dependence also exists be-
tween the converter dc-link voltage and the control signal In (40) is the dc-link voltage refer-
, manifested by current which depends from and as ence, and constants and are used for pole placement of
shown by this graph and (19). However, in current itself lies a the corresponding transfer function between and its ref-
hidden dependence of from given that is already a func- erence for a given settling time . These constants
tion of this control signal as shown in (18). This implies that the are determined per the ITAE criterion optimized for a step input
dependence between and is of second order and not first, as shown below [30]
the reason why a converter representation other than (19) is re-
quired to linearize the dc-link voltage dynamics.
A suitable alternative to model these dynamics is (21), which (41)
by considering the converter input–output power balance appro-
priately captures the above-mentioned subtlety, solely modeling Given that this is an outer loop there is no need for tracking ca-
the dependence between and as illustrated in Fig. 3(b). pacity, therefore, only an integrator has been added for possible
Notice that, given the cascaded control loop structure of this parameter mismatch compensation. Fig. 4(b) shows the block
scheme, current may be readily replaced by , namely the diagram of the dc-link voltage control loop.
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 369

Fig. 5. Complex signal flow graph of paralleled PWM-VSR representation. This graph captures the input current i and dc-link voltage v interaction as well as
the input–output converter power balance.

D. Reactive Power Loop converter representation. Particularly, this approach models the
The PWM-VSR reactive power is given by power flow passing through the converter instead of the dc cur-
rent generation. Consequently, the dc-link voltage results
from the instantaneous input–output power difference between
(42) the power system and the inverter load. This voltage signal flows
back then closing the nonlinear dynamic interaction of the con-
an expression that may be reduced to verter. Notice that in this case there is no appearance of ,
thus implying that may be directly controlled by together
(43) with the input current .

if is appropriately nulled by the space vector transforma- B. Input Current and DC-Link Voltage Loops
tion. Given the already linear dynamics of this relation, a simple Fig. 5 elicits the direct dependence of the PWM-VSR input
proportional–integral (PI) controller is used to regulate the con- current and dc-link voltage from its modulating signal ,
verter reactive power loop. This controller has the following being of first order for the former and of second order for the
structure: latter. These dependences are, respectively, observed through
the direct feed of product to the inductors block, and
(44) by the direct excitation of the capacitor block by current , a
function of as already revealed. This implies that both cur-
where is the reactive power reference, and rents and voltages may be controlled in parallel by the converter
the values of and are determined to ensure 20 dB at the modulating signal . However, only two variables may be con-
cutoff frequency of the current loop. These values are given by trolled by the complex space vector (equal to ), and
there are three distinct variables to control, specifically, the -
and -axes components of and itself. The solution nonethe-
(45)
less lies at hand.
Two key features to attain with any AFE converter based on
where is the nominal value of and is the current PWM-VSRs are unity input power factor, which implies con-
loop settling time. The block diagram of this control loop is also trolling the -axis current to zero, and, of course, regulation of
depicted in Fig. 4(b). the converter dc-link voltage. These, therefore, represent the two
main control objectives for this type of converter, which conve-
VI. PARALLELED REPRESENTATION AND CONTROL niently have the added advantage of inherently regulating the
OF THE PWM-VSR -axis current. This is so given that it is this current compo-
nent that actually generates the active power that regulates
A. Complex Signal Flow Graph
and, hence, by controlling this voltage is indirectly controlled,
By combining (18) and (21) a parallel current–voltage rep- becoming a bounded and stable internal dynamic of the con-
resentation of the PWM-VSR may be obtained. Fig. 5 shows trol system [10], [13]. Consequently, the reactive current and
the correspondent complex signal flow graph. As expected, both dc-link voltage may be readily chosen as control system out-
current/voltage interactions and converter power balance are ef- puts. The input transformation to replace by a new input in
fectively captured in this graph. Just as in the cascaded represen- order to attain IOL may, hence, be obtained as follows.
tation, the inputs to the system are the power supply voltage , As suggested by the previous analysis, a first-order differen-
the inverter load current , and the control input . The dy- tial relationship may be built for
namics of the reactive elements are also modeled by first-order
complex time delays of time constants and , where only the
capacitor block has been modified to accommodate the different (46)
370 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

Fig. 6. Block diagram of the paralleled control scheme for the PWM-VSR. (a) Inner q -axis current i and dc-link voltage v loops. (b) Outer reactive power q
loop.

In (46) is the -axis component of . Then, by observing by equating (49) and (50), which renders the following control
in Fig. 5 the terms to be cancelled out and equating (46) to the law:
-axis or imaginary part of (18), the input transformation for the
control of may be determined. This control law is given below

(47)

with defined in terms of the current error as follows:

(51)
(48)
where is defined in terms of the dc-link voltage error as
In (48) , and constants and are used for follows:
pole placement of the desired transfer function between
and . Their respective values are obtained per (37). This (52)
controller has parameter mismatch compensation as well as ex-
ponentially convergent tracking dynamics, which enhances the In (52) , and constants , and are
performance of its outer reactive power loop. used for pole placement of the desired third-order transfer func-
Now, for controlling the dc-link voltage a second-order differ- tion between and . Their respective values are
ential relation may be built according to the preceding analysis. obtained per the ITAE criterion optimized for a step input for a
For instance given settling time as shown below [30]. Fig. 6(a) shows
the resulting block diagram of this control scheme
(49)

where is the -axis component of . Correspondingly, dif-

ferentiation of (21) yields the second-order derivative of , (53)
explicitly shown below
Finally, (47) and (51) may be rewritten together paralleling
the control law structure presented in (7). This is shown below
with and defined as obvious

(50) (54)

This equation features an term, which proves the initial as-

sumption of a second-order dependence between and . C. Reactive Power Loop
The input transformation for the control of is then deter- The reactive power dynamics of the PWM-VSR are intrinsic
mined by direct inspection of the complex signal flow graph and to its topology, not to the control scheme chosen. Therefore,
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 371

Fig. 7. Complex signal flow graph of the cascaded PWM-CSR representation. (a) Input current i and dc-link current i dynamics. (b) Converter input–output
power balance-based dc-link current dynamics.

under the paralleled control scheme is in effect governed by as for the inductors, the self-induced negative feedback is only
(43), which naturally implies that an outer control loop may be observable from a nonstationary frame, also representing the
readily implemented to regulate it. This is possible given that cross-coupled nature of the - and -axes dynamics. Finally,
the only assumption taken in the derivation of (54) is that the dc-link inductor block is excited by the converter output
is zero, effectively achieved by proper synchronization of the voltage which is a function of and as shown in (31). Its
space-vector transformation. Then, a simple PI controller is im- output is the dc-link current which flows back through the
plemented given the aforementioned linear dynamics of . Its converter closing the state-variables interaction.
structure and parameters are also given by (44)–(45). The block Similarly to the PWM-VSR, the input–output power balance
diagram of this particular control loop is shown in Fig. 6(b). of the PWM-CSR cannot be directly studied from its current-
voltage representation. A separate signal flow graph is, hence,
developed and shown in Fig. 7(b). In this diagram, equivalent
VII. PWM-CSR PROPOSED CONTROL SCHEME to (32), the dc-link current dynamics are also represented by a
first-order complex time delay of time constant . Its inputs
A. Complex Signal Flow Graph are the converter input power, obtained through the product of
and , and voltage representing the converter output
Fig. 7(a) shows the signal flow diagram of the PWM-CSR
power. Hence, the instantaneous power difference between these
equivalent to its complex state-space model described by
increases or decreases the dc-link current. Notice that in this
(29)–(31). It is comprised of four blocks, namely, inductors and
graph current has also been replaced by , which becomes
capacitors that form a second-order input filter, dc-link inductor,
the control signal in this system.
and converter. The – filter dynamics are represented by the
series connection of two first-order complex time delays of time
constants and , respectively, whereas the dc-link inductor B. Inner Input-Current Loop
is modeled by a single first-order time delay of time constant
. The inputs to the system are the power supply voltage , Inspection of Fig. 7(a) clearly shows a second-order depen-
the current-source inverter voltage , and the control input dence of the converter input current from its control signal ,
, which, respectively, excite the input filter inductors, the as this current depends on the capacitors voltage , which in turn
dc-link inductor, and the converter itself. The inductors block depends on . Consequently, a second-order linear differential
is also excited by a negative, cross-coupled current feedback relationship may be constructed between and a new set of con-
or inductors back EMF, and by the input filter ac voltages . trol inputs with which is later replaced. Such a relationship
The output of this block, current , corresponds to the main is shown below
excitation to the ac capacitors, which are also excited by a
self-induced negative feedback voltage, and by the converter
ac current that captures the system nonlinearity. Just (55)
372 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

Fig. 8. Block diagram of the cascaded control scheme for the PWM-CSR. (a) Input current i loop. (b) Outer dc-link current i and reactive power q loops.

The second-order derivative of is then found by differentiating per the ITAE criterion optimized for a ramp input for a given
(29), which as depicted below presents an term that verifies a settling time as shown below [30]
the second-order dependence assumption previously made
(59)

This inner loop controller also presents parameter mismatch

compensation as well as tracking capacity provided by the right-
most term in (58). Fig. 8(a) shows the block diagram of this inner
current loop.

C. Outer DC-Link Current Loop

(56)
The inner input current loop used all the available control
variables that may be regulated by , hence, an alternative con-
Then, by further inspection of the converter complex signal flow verter representation is required to control its dc-link current.
graph and by equating (55) and (56) the input transformation for Conveniently, the PWM-CSR input–output power balance rep-
the control of current is determined. This control law is shown resentation given in (32) has no direct reference to , as op-
below posed to (31), and has as control signal the very input current
reference . Fig. 7(b) depicts the complex signal flow graph of
this converter model, clearly showing a first-order dependence
between the dc-link current and the control signal . Then,
the following linear differential relationship may be constructed
between and a new control input :

(57) (60)

The dc-link current control law is then obtained by equating (32)

With the new control input defined in terms of the current error
and (60), and giving heed to the signals and terms to be cancelled
as follows:
out using the complex signal flow graph. This transformation is
given below
(58)

In (58) , and . Constants (61)

, and are used to set the desired dynamic response for Just as for the PWM-VSR, is nulled through synchroniza-
this closed-loop system. Their respective values are obtained tion of the space-vector transformation. The dc-link current is,
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 373

Fig. 9. Experimental waveforms of the PWM-VSR under the cascaded control scheme. (a) AC waveforms operating at unity power factor and 80% loading.
(b) DC waveforms under same conditions. (c) AC waveforms operating at 0.86 leading power factor and 65% loading. (d) DC waveforms for this condition.
(e) Transient response to a 30% reactive power step. (f) Transient response to 50% load shed. Waveforms shown are: power supply phase voltage v (100 V/div,
2 ms/diV), converter line voltage v (200 V/div, 2 ms/diV), dc-link voltage v (200 V/div, 2 ms/diV), dc-link current i (10 A/div, 2 ms/diV) load current i load
(10 A/div, 2 ms/diV), input current i (20 A/div, 2 ms/diV), d- and q -axes current components i and i (10 A/div, 5 ms/diV), and converter reactive power q and
its reference q (3.8 kvar/div, 5 ms/diV).

hence, solely controlled by the -axis component of . The VIII. PERFORMANCE EVALUATION
new control signal is defined as
A. Experimental Setup
In order to evaluate the performance of the control schemes
(62) presented for the PWM-VSR and PWM-CSR two 5-kVA exper-
imental prototypes were implemented. The converters were fed
from a 110-V/50-Hz three-phase power supply, and controlled
with , and the dc-link current reference. by means of a TMS320C32 DSP-based digital control system
Constants and are used to set the desired dynamic response operating at a sampling frequency of 3 kHz. The input filter
for this loop, respectively determined by (41). The resulting inductance of the PWM-VSR employed was 1.4 mH, and its
control loop block diagram is depicted in Fig. 8(b). dc-link capacitance 2 mF. The cascaded control scheme was de-
signed to attain an input current settling time of 6.6 ms, and a
dc-link voltage settling time of 66 ms. The controller constants
D. Reactive Power Loop employed were and for the
current loop, and for the dc-link
The reactive power loop of this converter has the same struc- voltage loop, and and for the reactive power
ture of that employed by both PWM-VSR control schemes, loop. The parallel control scheme had the same current settling
given that the PWM-CSR reactive power is also defined by time, but a faster dc-link voltage settling time of 46.2 ms. The
(43). Accordingly, the controller employed for this control controller constants employed were and
loop is also (44)–(45). The corresponding control loop block for the -axis current loop, and
diagram is shown in Fig. 8(b). , and for the dc-link voltage loop.
374 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

Fig. 10. Experimental waveforms of the PWM-VSR under the paralleled control scheme. (a) AC waveforms operating at unity power factor and 75% loading.
(b) DC waveforms under same conditions. (c) AC waveforms operating at 0.86 leading power factor and 65% loading. (d) DC waveforms for this condition.
(e) Transient response to a 30% reactive power step. (f) Transient response to 50% load shed. Waveforms shown are: power supply phase voltage v (100 V/div,
2 ms/diV), converter line voltage v (200 V/div, 2 ms/diV), dc-link voltage v (200 V/div, 2 ms/diV), dc-link current i (10 A/div, 2 ms/diV) load current i load
(10 A/div, 2 ms/diV), input current i (20 A/div, 2 ms/diV), d- and q -axes current components i and i (10 A/div, 5 ms/diV), and converter reactive power q and
its reference q (3.8 kvar/div, 5 ms/diV).

Its reactive power loop employed and . For phase voltage , converter line voltage , dc-link voltage
the PWM-CSR the input filter inductance and capacitance were , dc-link current , load current , input current -
respectively 1.4 mH and 90 F, while its dc-link inductance was and -axes current components and , and the converter
of 40 mH. This control scheme was designed to meet a current reactive power and its reference . These waveforms show
settling time of 6.6 ms, and a dc-link current settling time of 66 the good results attained with this control scheme, presenting
ms. The controller gains were , linear dynamics and a virtually nonexistent coupling between
and for the input current loop, the - and -axes channels.
and for the dc-link current loop, and and Fig. 10 shows experimental waveforms of the PWM-VSR
for the reactive power loop. under the paralleled control scheme. Fig. 10(a) and (b) shows
steady-state waveforms at unity input power factor, Fig. 10(c)
B. Results and (d) the operation at 0.86 leading power factor, while
Fig. 9 shows steady-state and transient waveforms of the Fig. 10(e) and (f) shows the converter transient response when
PWM-VSR operating with the cascaded control scheme. subject to a reactive power step command of 30%, and a
Fig. 9(a) and (b) depicts the converter operating at unity input 50% load rejection respectively. The advantage of this control
power factor and 80% nominal power, and Fig. 9(c) and (d) the scheme lies in a faster dc-link voltage control when compared
operation at 0.86 leading power factor under 65% power de- to the cascaded approach. However, due to the intrinsic dif-
mand. Fig. 9(e) on the other hand shows the transient response ferent time response of the -axis current and dc-link voltage
of the converter when subject to a reactive power step command control loops a slight coupling occurs between the - and the
of 30%, and Fig. 9(f) the response to a 50% load rejection. -axis currents as shown in Fig. 10(e). Nonetheless, this does
The waveforms shown in these pictures are: power supply not impair the converter capability to operate as a reactive
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 375

Fig. 11. Experimental waveforms of the PWM-CSR under cascaded control scheme. (a) AC waveforms operating at unity power factor and 80% loading. (b) DC
waveforms under same conditions. (c) AC waveforms operating at 0.8 leading power factor and 80% loading. (d) DC waveforms for this condition. (e) Transient
response to a 30% reactive power step. (f) Transient response to 25% dc-link current step. Waveforms shown are: power supply phase voltage v (100 V/div,
2 ms/div), input filter phase voltage v (100 V/div, 2 ms/div), dc-link voltage v (100 V/div, 2 ms/div), load voltage v (100 V/div, 2 ms/div), converter line
current i (20 A/div, 2 ms/div), dc-link current i (20 A/div, 2 ms/div), input current i (20 A/div, 2 ms/div), d- and q -axes current components i and i (10 A/div,
5 ms/div), and converter reactive power q and its reference q (3.8 kvar/div, 5 ms/div).

power compensator as shown by Fig. 10(c)–(e). This inherent scheme, featuring linear converter dynamics with decoupled -
capacity had not been exploited nor explored by previous works and -axes channels.
[12], [13], [26], [27]. The waveforms shown in these figures The – -axes waveforms depicted in Figs. 9–11 show an
are: power supply phase voltage , converter line voltage , apparent oscillation at the sixth harmonic (300 Hz), which
dc-link voltage , dc-link current , load current , input may be easily observed on the PWM-CSR and currents
current - and -axes current components and , and the due to the presence of the – filter. This is not a system
converter reactive power and its reference . resonance but the result of the modulation that takes place when
Fig. 11 shows steady-state and transient waveforms of the converting ac waveforms into the – synchronous frame. These
PWM-CSR operating under the cascaded control scheme pre- ac waveforms contain fifth and seventh harmonics—generated
sented. Fig. 11(a) and (b) depicts the converter operating at unity by the space-vector modulator regular sampling process—and,
input power factor and 80% nominal power, and Fig. 11(c) and thus, generate sixth harmonics on the converted – variables. To
(d) the operation at 0.8 leading power factor under the same verify this, a 100-kW PWM-CSR was simulated and subjected to
power conditions. The transient response of this converter is a 45% step change in its reactive power reference (Fig. 12). The
shown in Fig. 11(e) and (f), which, respectively, depict a reac- much higher power rating and lower damping of this converter
tive power step command of 35% and a 25% dc-link current system should have excited the sixth harmonic oscillation had
step. The waveforms shown are: power supply phase voltage it been a system resonance. However, as shown in Fig. 12, the
, converter line current , dc-link voltage , dc-link current converter showed linear and decoupled dynamics without any
, load voltage , input current - and -axes current resonance problems. The 300-Hz oscillation is present as well,
components and , and the converter reactive power and but at the same level shown in the experimental waveforms,
its reference . These results validate the presented control which verifies its modulation-generated nature.
376 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 52, NO. 2, APRIL 2005

Fig. 12. Simulated waveforms of a 220-V/50-Hz 100-kW PWM-CSR under cascaded control scheme subject to a 45% reactive power step. Waveforms shown
are: power supply phase voltage v , input filter phase voltage v , input current i, converter line current i , current i d-q axes components i and i , and converter
reactive power q and its reference q .

IX. CONCLUSION [8] J. Choi and S. Sul, “New current control concept-minimum time current
control in the three-phase PWM converter,” IEEE Trans. Power Elec-
This paper has presented the use of complex state variables tron., vol. 12, no. 1, pp. 124–131, Jan. 1997.
to model and control AFE converters. Particularly, three-phase [9] H. Mao, D. Boroyevich, and F. Lee, “Novel reduced-order small-signal
model of a three-phase PWM rectifier and its application in control de-
PWM-VSRs and PWM-CSRs were analyzed with the aid of the sign and analysis,” IEEE Trans. Power Electron., vol. 13, no. 3, pp.
graphic capabilities of this approach, i.e., complex signal flow 511–521, May 1998.
graphs. With them, high-performance control laws were directly [10] J. Slotine and W. Li, Applied Nonlinear Control. Englewood Cliffs,
NJ: Prentice-Hall, 1991, pp. 207–275.
and intuitively obtained based on IOL principles. Moreover, this [11] P. Rioual, H. Pouliquen, and J. Louis, “Nonlinear control of PWM rec-
was attained while successfully shunning the well known math- tifier by state feedback linearization and exact PWM control,” in Proc.
ematical complexity of this nonlinear control strategy. Conse- IEEE PESC’94, vol. 2, 1994, pp. 1095–1102.
[12] D. Lee, G. Lee, and K. Lee, “DC-bus voltage control of three-phase
quently, linear and decoupled dynamics were attained by both AC/DC PWM converters using feedback linearization,” IEEE Trans.
cascaded and paralleled control schemes of the PWM-VSR, and Ind. Appl., vol. 36, no. 3, pp. 826–833, May/June 2000.
also by the cascaded scheme of the PWM-CSR. Moreover, these [13] T. Lee, “Input-output linearization and zero-dynamics control of three-
control schemes featured enforced stability obtained through the phase AC/DC voltage-source converters,” IEEE Trans. Power Electron.,
vol. 18, no. 1, pp. 11–22, Jan. 2003.
respective linearization of their dc-link voltage and current dy- [14] J. Espinoza, G. Joos, M. Perez, and L. Moran, “Stability issues in three-
namics. Finally, experimental results from 5-kVA TMS320C32 phase PWM current/voltage source rectifiers in the regeneration mode,”
DSP-based laboratory prototypes verified the excellent perfor- in Proc. IEEE ISIE’00, vol. 2, 2000, pp. 453–458.
[15] W. Wang and C. Wang, “Speed and efficiency control of an induction
mance attained by these AFE converters. motor with input-output linearization,” IEEE Trans. Energy Convers.,
vol. 14, no. 3, pp. 373–378, Sep. 1999.
[16] J. Jung, S. Lim, and K. Nam, “A feedback linearizing control scheme
REFERENCES for a PWM converter-inverter having a very small DC-link capacitor,”
[1] W. Sweet and E. Bretz, “Restructuring the thin-stretched [US electricity IEEE Trans. Ind. Appl., vol. 35, no. 5, pp. 1124–1131, Sep./Oct. 1999.
supply industry],” IEEE Spectr., vol. 37, pp. 43–49, Jun. 2000. [17] T. Bukas and T. Habetler, “High-performance induction motor speed
[2] S. Bernet, “Recent developments of high power converters for industry control using exact feedback linearization with state and state derivative
and traction applications,” IEEE Trans. Power Electron., vol. 15, no. 6, feedback,” in Proc. IEEE PESC’03, vol. 1, 2003, pp. 15–24.
pp. 1102–1117, Nov. 2000. [18] Z. Yao, M. Dawande, and V. Rajagopalan, “Controller design for ad-
[3] V. Blasko and V. Kaura, “A new mathematical model and control of a vanced reactive power compensators based on input-output lineariza-
three-phase ac-dc voltage source converter,” IEEE Trans. Power Elec- tion,” in Proc. IEEE PESC’97, vol. 2, 1997, pp. 936–941.
tron., vol. 12, no. 1, pp. 116–123, Jan. 1997. [19] H. Kanaan, K. Al-Haddad, R. Chaffai, L. Duguay, and F. Fnaiech,
[4] J. Rodriguez, J. Lai, and F. Peng, “Multilevel inverters: A survey of “Input-output feedback linearization technique applied to a CITPR,” in
topologies, controls, and applications,” IEEE Trans. Ind. Electron., vol. Proc. IEEE PESC’01, vol. 3, 2001, pp. 1321–1326.
49, no. 4, pp. 724–738, Aug. 2002. [20] P. Krause, O. Wasynczuk, and S. Sudhoff, Analysis of Electric Ma-
[5] J. Espinoza and G. Joos, “State variable decoupling and power flow con- chinery and Drive Systems, 2nd ed. New York: Wiley/IEEE Press,
trol in PWM current-source rectifiers,” IEEE Trans. Ind. Electron., vol. 2002.
45, no. 1, pp. 78–87, Feb. 1998. [21] J. Holtz, “The representation of AC machine dynamics by complex
[6] Y. Xiao, B. Wu, F. DeWinter, and R. Sotudeh, “A dual GTO current- signal flow graphs,” IEEE Trans. Ind. Electron., vol. 42, no. 3, pp.
source converter topology with sinusoidal input for high-power appli- 263–271, Jun. 1995.
cations,” IEEE Trans. Ind. Appl., vol. 34, no. 4, pp. 878–884, Jul./Aug. [22] , “On the spatial propagation of transient magnetic fields in AC
1998. machines,” IEEE Trans. Ind. Appl., vol. 32, no. 4, pp. 927–937, Jul./Aug.
[7] S. Hiti, V. Vlatkovic, D. Boroyevich, and F. Lee, “A new control algo- 1996.
rithm for three-phase PWM buck rectifier with input displacement factor [23] J. Jiang and J. Holtz, “High dynamic speed sensorless AC drive with
compensation,” IEEE Trans. Power Electron., vol. 9, no. 2, pp. 173–180, on-line model parameter tuning for steady-state accuracy,” IEEE Trans.
Mar. 1994. Ind. Electrron., vol. 44, no. 2, pp. 240–246, Apr. 1997.
BURGOS et al.: COMPLEX STATE-SPACE MODELING AND NONLINEAR CONTROL OF AFE CONVERTERS 377

[24] J. Holtz, “Sensorless position control of induction motors-An emerging Eduardo P. Wiechmann (S’81–M’86–SM’94)
technology,” IEEE Trans. Ind. Electron., vol. 45, no. 6, pp. 840–852, received the Electronics Engineering degree from
Dec. 1998. Santa Maria University, Valparaiso, Chile, in 1975,
[25] , “Sensorless control of induction motor drives,” Proc. IEEE., vol. and the Ph.D. degree from Concordia University,
90, no. 8, pp. 1359–1394, Aug. 2002. Montreal, QC, Canada, in 1985.
[26] H. Sugimoto, S. Morimoto, and M. Yano, “A high-performance con- Since 1976, has been with the University of Con-
trol method of a voltage-type PWM converter,” in Proc. IEEE PESC’88, cepción, Concepción, Chile, where he is currently
1988, pp. 360–368. a Professor in the Electrical Engineering Depart-
[27] S. Fukuda Y. Iwaji and T. Aoyama, “Modeling and control of sinusoidal ment. His research interests are power converters,
PWM rectifiers,” in Proc. EPE’93, vol. 4, 1993, pp. 115–120. high-current rectifiers, ac drives, UPS systems, har-
[28] E. Wiechmann, R. Burgos, and J. Holtz, “Active front-end converter for monics, and power-factor control in industrial power
medium-voltage current-source drives using sequential-sampling syn- distribution systems. His industrial experience includes more than 6000 hours
chronous space vector modulation,” IEEE Trans. Ind. Electron., vol. 50, working on engineering projects and consulting. He has published numerous
no. 6, pp. 1275–1289, Dec. 2003. technical papers and has coauthored technical books.
[29] E. Wiechmann, P. Ziogas, and V. Stefanovic, “Generalized functional Dr. Wiechmann is currently Chairman of the IEEE Chilean Joint Chapter
model for three-phase PWM inverter/rectifier converters,” IEEE Trans. of the IEEE Industry Applications, IEEE Power Electronics, and IEEE Indus-
Ind. Appl., vol. 23, no. 2, pp. 236–246, Mar./Apr. 1987. trial Electronics Societies. He is the recipient of the year 2000 Concepción City
[30] R. Dorf, Modern Control Systems. Reading, MA: Addison-Wesley, Award for Outstanding Achievements in Applied Research.
1992.

Joachim Holtz (M’87–SM’88–F’93) graduated in

1967 and received the Ph.D. degree in 1969 from the
Technical University Braunschweig, Braunschweig,
Germany.
In 1969, he became an Associate Professor and, in
1971, he became a Full Professor and Head of the
Control Engineering Laboratory, Indian Institute of
Technology, Madras, India. In 1972, he joined the
Siemens Research Laboratories, Erlangen, Germany.
From 1976 to 1998, he was a Professor and Head of
the Electrical Machines and Drives Laboratory, Wup-
Rolando P. Burgos (S’96–M’03) received the the pertal University, Wuppertal, Germany. He is currently Professor Emeritus and
Electronics Engineering, M.S., and Ph.D. degrees a Consultant. He has published extensively, including 11 invited papers in jour-
with honors from the University of Concepción, Con- nals. He is the coauthor of four books and the holder of 30 patents.
cepción, Chile in 1997, 1999, and 2002 respectively. Dr. Holtz was the recipient of the IEEE Industrial Electronics Society Dr.
In 2002, he joined as a Postdoctoral Fellow the Eugene Mittelmann Achievement Award, the IEEE Industry Applications So-
Center for Power Electronics Systems, Virginia ciety Outstanding Achievement Award, the IEEE Power Electronics Society
Polytechnic Institute and State University, Blacks- William E. Newell Field Award, the IEEE Third Millennium Medal, and the
burg, where, since 2003, he has been a Research IEEE Lamme Gold Medal. He has earned six IEEE Prize Paper Awards. He is
Scientist. His areas of interest are the modeling and Past Editor-in-Chief of the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS.
control of three-phase static power converters. He has received nine Prize Paper Awards.